Opinions are my own – For my predictions for the previous year 2021 please click here

For the year of 2022 – 2023:

1-      Deep Learning models training will be king in 2022 in Machine Learning. The goal is to generate human-like semantic new data (Texts, Images and Code). In 2021 we have trained large ML models with billions of parameters (OpenAI GPT-3, Turing, codex etc). In 2022 we will continue training models with more parameters, but the difference now will be to reduce the bias/stereotypes in the new content that is created.

a.       One key trend is to allow users to fine tune their models, tailor to their specific domain this will lead higher quality outputs

b.       Another trend is the increase of data annotation/labeling tools and vendors to tag data with more inclusive labels.

c.       Finally, there will be much focus on digital signature to certify the new data that is being generated by the publisher to reduce propaganda

2-     The Metaverse in IOT space will have the biggest impact: this means in simple term representing the physical world by building a digital replica. We will continue to see multiple approaches for the Metaverse in 2022.

a.        For some they will focus on Mixed Reality and/or Gaming.  Others will focus on industry solutions such as Autonomous driving/Asset performance/Supply chain. The key is to monitor how these different companies will use personal data that is being collected by their solutions.  

b.       The cloud will be the oxygen for the metaverse in terms of elasticity and scalability. The cloud is where companies will create the digital twins (copies of your physical environment and synchronize the data in near-real-time). You use the cloud to run your intelligence at scale and extend the intelligence back to the edge. The cloud will not replace the edge.

3-     Polyengines will be picking up greatly in 2022. At a high level this means abstracting and democratizing the type of compute that will be used (SQL, Spark) to process user queries; The concept behind polyengines will allow the platform to automatically process your query and pick the best compute needed. The user is only responsible for the query. They should not worry about running their query in this or that platform anymore!

a.       You write your code in your language of choice (python/SQL) then the platform uses the best engine independently SQL and/or Spark to process your query based on performance/price.

b.       Today, users must support multiple platforms (Analytical system, Operational System, Data Warehousing, etc) this requires a lot of skills and increased cost. With polyengines let the magic happen. All you worry about is your code. Everything else is managed by the platform.

4-     Reduce Data Movement and data duplication to enable real-time analytics and 360 view of the data across the enterprise will have much focus in 2022.

a.       There are currently, multiple new data patterns such as data mesh/data fabric. The key is to give back power to the business users. IT will be responsible for observability of the data.

b.       No more data silos moving forward Data will represent specific domains/products (Marketing, HR etc.). Each business unit will be using the same type of structure/tools across the organization. The difference now, they will be following the same standards at the enterprise level. This will enable in-place data sharing and exploration of the data across the different business units without the need of a central repository aka data lake.

c.       Reverse ETL will be king! Bring the intelligence back to the application.

5-     Companies cannot be anymore retail or supply chain or manufacturing or health and other types of companies they must all be first AI companies. This is not a new prediction, but it is needed for any company to survive moving forward in 2022 and beyond. This is the only way they can continue growing.

a.        The companies will need to incorporate at every level of their business automation and intelligence.

b.       They will need to process and collect all the digital data available internally and externally to reduce costs, improve speed of delivery, and understand their customers top priorities.

c.       The key is to decentralize and organize themselves as Domain/Product solution.

Bonus:

6-     Privacy and Identity security is going to become even of bigger concerns with the rise of AI and metaverse solutions.

a.       Identity today has many problems because our information is mainly stored on a centralized server, it becomes a target for hackers to steal our information. To overcome this there will be an increase in Blockchain solution to provide a unique ID for each user to enable to share their own information

b.       An “AI bill of Rights” will be required with the rise of the metaverse. Imagine now third parties can have access to everything about us, our location, spending habits, driving patterns with automation, our reactions on any physical reaction. The solution here is not technical but more bureaucratic.